Running Head: CLARIFYING PSYCHOPATHY CONCEPTIONS 1. Clarifying Conceptions Underlying Adult Psychopathy Measures: A Construct Validity Metric Approach

Transcription

1 Running Head: CLARIFYING PSYCHOPATHY CONCEPTIONS 1 Clarifying Conceptions Underlying Adult Psychopathy Measures: A Construct Validity Metric Approach Patrick J. Kennealy Travis County Community Justice Services Jennifer L. Skeem University of California, Berkeley Scott O. Lilienfeld Emory University Acknowledgments: The authors would like to thank Christopher J. Patrick for his substantial feedback on this study. We also acknowledge and appreciate the assistance and cooperation of the following agencies in collecting data for this research, however, none of the opinions or conclusions expressed in this article reflect any official policy or position of any of these institutions: Drug Abuse Comprehensive Coordinating Office (DACCO), Tampa, FL; Florida Department of Corrections; Gateway Foundation, Huntsville, TX; Nevada Department of Prisons; Odyssey House, Salt Lake City, UT; Operation PAR, Pinellas Park, FL; Oregon Department of Corrections; Texas Department of Criminal Justice Institutional Division; Utah Department of Corrections; Volunteers of America, Portland, OR; Westcare, NV. This study was funded by the National Institute of Mental Health (PI: Norman Poythress; Grant #R01 MH A1). Contact: Patrick J. Kennealy, Ph.D. Criminal Justice Researcher Travis County Community Justice Services 411 West 13 th Street Austin, TX pjkennealy@gmail.com

2 CLARIFYING PSYCHOPATHY CONCEPTIONS 2 Abstract The two most widely used measures of psychopathy the Psychopathy Checklist-Revised (PCL- R; Hare, 2003) and Psychopathic Personality Inventory (PPI; Lilienfeld & Andrews, 1996) ostensibly measure the same construct, but seem to emphasize different conceptualizations of psychopathy. To clarify distinctions between these two measures, this study investigates how well the PCL-R and PPI map onto historical conceptualizations of psychopathy. In a sample of 1,281 adult male offenders, we used construct validity metrics (Westen & Rosenthal, 2003) to investigate the degree to which the relations between psychopathy measures and 14 theoretically meaningful criterion variables are consistent with predictions of researchers familiar with historical conceptualizations of psychopathy. This study s primary findings highlight both similarities and differences between psychopathy measures. At the total score level, both the PCL-R and PPI are most consistent with the lovelessness-based conceptualization of the McCords (1964) and the affective dysfunction-centered conceptualization of Karpman s (1948a) secondary psychopathy. Although key similarities between these measures also emerge at the factor level, the PPI generally demonstrated higher levels of consistency between theory-based predictions and observed relations with conceptualization-relevant criterion measures than did the PCL-R. These findings suggest that these psychopathy measures are most consistent with secondary psychopathy conceptualizations, offering important implications for practice (e.g., treatability and risk assessment) and research (e.g., development and refinement of measures). Keywords: psychopathy, Psychopathy Checklist-Revised, Psychopathic Personality Inventory, construct validity

3 CLARIFYING PSYCHOPATHY CONCEPTIONS 3 Clarifying Conceptions Underlying Adult Psychopathy Measures: A Construct Validity Metric Approach Psychopathy is a personality disorder characterized by a complex mixture of interpersonal, affective, and behavioral traits, including superficial charm, narcissism, lack of anxiety, shallow affect, lack of remorse, dishonesty, and inadequately motivated antisocial behavior (Cleckley, 1976). Measures of psychopathy have been commonly applied to inform legal decisions about justice-involved people that turn upon dangerousness and treatment amenability (Skeem & Cooke, 2010). In fact, surveys of forensic diplomates in the United States indicate that the Psychopathy Checklist-Revised (PCL-R; Hare, 2003), the best-known measure of psychopathy, is one of the instruments most commonly used to assess future risk of violence for the court (Archer, Buffington-Vollum, Stredny, & Handel, 2006; Lally, 2003; Tolman & Mullendore, 2003; Viljoen, McLachlan, & Vincent, 2010). Despite the construct s legal applications, the field largely lacks consensus on how the term psychopathy should be defined and operationalized. There are several conceptualizations of psychopathy, each of which highlights distinct characteristics and hypothesized etiologies (see Cleckley, 1941; Karpman, 1948a; Lykken, 1957; McCord & McCord, 1964; Mealey, 1995). There are also differences in operationalizations of the construct and spirited debates about which is superior. The two leading measures of psychopathy are the PCL-R and Psychopathic Personality Inventory (PPI; Lilienfeld & Andrews, 1996) and its revised version, the PPI-R (Lilienfeld & Widows, 2005). Some have criticized the PCL-R s emphasis on violence and other criminal behavior (Skeem & Cooke, 2010) and de-emphasis of explicit indicators of positive adjustment (e.g., good intelligence, absence of nervousness, and suicide rarely carried out; Patrick, 2006; see also Cleckley, 1976). Others have criticized the PPI s emphasis on

4 CLARIFYING PSYCHOPATHY CONCEPTIONS 4 fearlessness, contending that this trait is largely irrelevant to the condition (see Lilienfeld, Patrick, Benning, Berg, Sellbom, & Edens, 2012; Marcus, Fulton, & Edens, 2013; Miller & Lynam, 2012; Patrick, Venables, & Drislane, 2013; for differing viewpoints). As a result of this lack of consensus, many clinicians and researchers may be utilizing psychopathy measures to inform legal decisions under mistaken assumptions. For example, a professional may assume that an individual assessed as psychopathic is at risk for violent behavior, regardless of the psychopathy measure or specific scales used. The differences in content and predictive utility for future violence between the PCL-R and PPI make this possibility concerning. This is particularly salient as recommendations can result in longer and more severe sanctions in the criminal justice system. A promising way to address this ambiguity is to identify which of a number of well-articulated historical conceptualizations of psychopathy (Cleckley, 1941; Karpman, 1948a; Lykken, 1957; McCord & McCord, 1964; Mealey, 1995) is (or are) most consistent with each measure. This study is designed to identify conceptualizations that best match the PCL-R and PPI. Such an approach offers two advantages. First, when a measure is grounded in a theoretical conceptualization, it can provide guidance for refining the measure over time to minimize error and better approximate that conceptualization (Smith, 2005). Further, a precise theory can guide researchers in designing hypothesis-driven research that addresses important unanswered questions (Cronbach, 1988; Kane, 2001), including etiological factors and the role of criminal behavior in psychopathy. Second, a theory can shed light on the clinical implications (e.g., violence risk, treatability) of a measure (Cronbach, 1988). Clinicians tend to approach the diagnostic enterprise with prototypes of disorders in mind, so it may be crucial to ensure they are accessing the appropriate prototype or conceptualization when they use a particular measure

5 CLARIFYING PSYCHOPATHY CONCEPTIONS 5 (Cantor, Smith, French, & Mezzich, 1980; Genero & Cantor, 1987; Westen, Shedler, & Bradley, 2006). Because the psychopathy label can lead to adverse consequences including harsh sentences in the justice system (Edens, Desforges, Fernandez, & Palac, 2004; Kelley, Edens, Mowle, Penson, & Rulseh, 2018) it is crucial to inform professionals about the specific conceptualization of psychopathy assessed by each measure, so they can select a measure that is appropriate for informing a particular legal and/or clinical decision. Psychopathy Conceptualizations Although many people who use measures of psychopathy may think of the construct as a monolithic entity, theory and research evidence suggest otherwise. The field currently features two well-developed groups of conceptualizations: primary and secondary psychopathy. These conceptualizations can be differentiated based on their proposed characteristics and etiological assumptions (see Table 1). For the purposes of this discussion, each of these conceptualizations was selected based on their consistency with Strauss and Smith s (2009) definition of a precise theory, which requires a theory to identify (a) a relatively homogenous group of interest and (b) a causal process that characterizes the described group of interest. Primary Psychopathy. These conceptualizations emphasize affective deficits that are often assumed to largely reflect a genetic predisposition. Cleckley (1941; 1976) described psychopathic people as possessing impoverished affective reactions and believed that an absence of nervousness or anxiety typified the disorder. Similarly, Karpman s (1948a) primary psychopaths were characterized by a lack of conscience (Poythress & Skeem, 2006) and complex emotional reactions Karpman, 1946, p. 286). Rather than focusing on broad affective deficits, Lykken s (1957, 1995) primary psychopathy conceptualization focuses on a

6 CLARIFYING PSYCHOPATHY CONCEPTIONS 6 specific deficit in fear. Finally, Mealey s (1995) primary psychopath can be distinguished based on a paucity of such emotions as anxiety, shame and guilt. Secondary Psychopathy. In contrast, secondary conceptualizations typically highlight affective dysfunction thought to arise largely from environmental factors. Karpman s (1941; 1946; 1955) secondary psychopathy is characterized by highly disturbed emotional reactions or disorders that reflect exposure to early adversity (e.g., parental rejection, physical or sexual abuse). McCord and McCord s (1964) secondary psychopathy is characterized by guiltlessness and lovelessness thought to be influenced by early parental rejection and neglect (Patrick et al., 2010). Finally, Mealey s (1995) secondary psychopathy is thought to reflect below-average intelligence and environmental disadvantage, including low socioeconomic status, a...disrupted family life associated with parental neglect, abuse, inconsistent discipline, and the use of punishment as opposed to rewards (p. 533). Leading Psychopathy Measures These competing conceptualizations have exerted a direct influence on the field s two most widely used and researched measures of adult psychopathy, namely, the PCL-R and PPI. The Psychopathy Checklist (PCL; Hare, 1980) and the subsequent PCL-R were developed with prison populations in an effort to reliably assess psychopathy as inspired by the works of Cleckley, Karpman, the McCords, and others (Hare & Neumann, 2010). The resulting 20-item checklist is completed by a clinician after a detailed semi-structured interview and the review of relevant file information. Early factor analyses of the PCL suggested that the measure consisted of two moderately correlated dimensions: an Interpersonal-Affective Factor and a Social Deviance Factor (Hare, Harpur, Hakstian, Forth, Hart, & Newman, 1990; Harpur, Hare, & Hakstian, 1989). The Interpersonal-Affective factor measures characteristics such as superficial

7 CLARIFYING PSYCHOPATHY CONCEPTIONS 7 charm, interpersonal dominance, shallow affect, and callousness whereas the Social Deviance factor measures disinhibition and a wide range of antisocial and criminal behaviors (Hare, 2003). The PPI was designed as a self-report measure of psychopathic traits for use in noncriminal samples, although it has since been extended to noncriminal samples. Rather than basing the PPI on a specific conceptualization, Lilienfeld and Andrews (1996) wrote items that assessed 34 focal personality constructs that were identified as important to a range of psychopathy conceptualizations, including influential primary psychopathy conceptualizations, after an expansive literature review. Lilienfeld and Andrews focused explicitly on personality, excluding items that assessed criminal behavior. Using successive item-level factor analyses, the authors identified eight subscales that yield a total score. Similar to the PCL-R, subsequent exploratory factor analyses of the PPI s eight scales have identified a two-factor structure in community samples (Benning, Patrick, Hicks, Blonigen, & Krueger, 2003; cf., Neumann, Malterer, & Newman, 2008) and offender samples (Patrick, Edens, Poythress, Lilienfeld, & Benning, 2006). However, in contrast to those of the PCL-R, these factors are largely uncorrelated. The first factor (Fearless Dominance) assesses adventurousness, social dominance, and low levels of anxiety. The second factor (Self-Centered Impulsivity) assesses egocentricity, manipuativeness, disinhibition, and hostile attribution bias. Of the PPI s eight subscales, only the Coldheartedness subscale, which assesses a lack of affective empathy and callousness, was excluded from this two-factor structure because it did not load highly on either factor. Given the centrality of affective empathy and similar traits to conceptualizations of psychopathy, some researchers advocate the use of the PPI s Coldheartedness scale as a third standalone factor (Lilienfeld & Widows, 2005; Miller & Lynam, 2012; Neumann, Malterer, & Newman, 2008). Limited Overlap between Measures

8 CLARIFYING PSYCHOPATHY CONCEPTIONS 8 Although they both measure psychopathy, there are two key differences between PCL- R and PPI operationalizations. First, in a departure from its Cleckleyan (1976) roots, the PCL-R excludes explicit indicators of positive adjustment (e.g., good intelligence, absence of nervousness, and suicide rarely carried out). Patrick (2006) speculated that the omission of positive adjustment indicators could have resulted from efforts to improve reliability during the measure s development. Specifically, Patrick (2006) posited that because the majority of Cleckley s psychopathy criteria (12 of 16) reflected maladjustment, the criteria gauging positive adjustment (4 of 16) may have been dropped from the measure because they did not fit with the majority of items. In contrast, the PPI includes Stress Immunity and Social Potency subscales, which detect lack of trait anxiety and interpersonal dominance/self-confidence, respectively. Second, in another departure from Cleckley (1976), the PCL-R includes a number of items that explicitly gauge criminal behavior, and other items that can be scored based on the presence of criminal acts. For example, callousness may be inferred from the offender's discussion of crimes and victims and from file reports of violent behavior (Hare, 2003s). This inclusion of criminal behavior could be attributable to the fact that the PCL-R was developed with samples of offenders. Items assessing the extent of an individual s criminal history may have been added to better (a) distinguish more severe from less severe offenders or (b) identify criminal expressions of psychopathic traits (Patrick, 2006). In contrast, the PPI explicitly does not include items explicitly assessing antisocial or criminal behavior, in an effort to unconfound the personality dispositions ostensibly underpinning psychopathy from their associated antisocial actions (Lilienfeld, 1998). Because of these differences, the PCL-R and PPI may identify overlapping but separable constructs (Lilienfeld, 1998; Marcus et al., 2013). Indeed, correlational evidence suggests that at

9 CLARIFYING PSYCHOPATHY CONCEPTIONS 9 the total score level, the two measures explain less than 30% of each other s variance (Poythress, Edens & Lilienfeld, 1998). This concern is highlighted by Marcus and colleagues (2013) metaanalytic findings (n = 5,432) on the correspondence between PCL-R and PPI factors. Specifically, the PCL-R s Interpersonal-Affective factor is modestly associated with the PPI s Fearless Dominance (r =.21) and Self-Centered Impulsivity (r =.20) factors. In contrast, the PCL-R s Social Deviance factor is weakly associated with the PPI s Fearless Dominance (r =.15) but moderately associated with the PPI s Self-Centered Impulsivity (r =.41) factors. Hence, individuals who score highly on the PCL-R may be quite different from those who score highly on the PPI, and vice-versa. Construct Validity Coefficients: A Way Forward One promising approach to address this lack of clarity is to identify which of several historical psychopathy conceptualizations (Cleckley, 1941; Karpman, 1948a; Lykken, 1957; McCord & McCord, 1964; Mealey, 1995) is (or are) best captured by a given psychopathy measure via Cronbach and Meehl s (1955) process of construct validation. In this process, the degree that a conceptualization is captured by a measure is evaluated based on how well the measure relates to other theoretically relevant variables. Traditionally, researchers have established construct validity by eye-balling the pattern of relations observed between a measure and criterion variables. Based on these impressions, researchers draw subjective inferences about whether the measure is working as it should or if there is a need for revisions in the measure, the theory, or both (Westen & Rosenthal, 2003). In a departure from this informal approach to ascertaining construct validity, Westen and Rosenthal (2003) developed fit metrics that quantitatively gauge how well a measure s observed correlations converge with theory-predicted correlations. The magnitude of these coefficients

10 CLARIFYING PSYCHOPATHY CONCEPTIONS 10 reflects the degree of match between the hypothesized and obtained pattern of correlations. Once calculated, these indices can be compared to identify which conceptualization is best captured by a specific measure. For example, this methodology allows one to test how well the PCL-R replicates Cleckley s (1976) theoretically expected pattern of associations with criterion variables compared with how well the PCL-R replicates expected patterns of associations dictated by competing theories of psychopathy (Karpman, 1948a; Lykken, 1957; McCord & McCord, 1964, Mealey, 1995). To date, few studies have used these construct validity metrics in relation to assessment tools (Bombel, Mihura, & Meyer, 2009; Kosson, Blackburn, Byrnes, Park, Logan, & Donnelly, 2008; Poythress, Lilienfeld, Skeem, Douglas, Edens, Epstein, & Patrick, 2010), and research has yet to use these metrics to map psychopathy measures onto conceptualizations. Nonetheless, one study did use Westen and Rosenthal s (2003) construct validity coefficients to gauge how well the PPI replicated the PCL-R s pattern of relations with a range of criterion measures (Poythress et al., 2010). Rather than make theory-predicted correlations, the authors used the PCL-R s observed correlations with criterion measures as the basis of comparison for the PPI s observed correlations. The results showed very high levels of convergence between the patterns of the PCL-R and PPI at the total score level (rcontrast-cv =.88). Further, the PPI Fearless Dominance scale demonstrated stronger convergence with the PCL-R Interpersonal-Affective factor (rcontrast-cv =.71) than the PCL-R Social Deviance factor (rcontrast- CV =.14), whereas the PPI Self-Centered Impulsivity scale was more consistent with the PCL-R Social Deviance factor (rcontrast-cv =.93) than the PCL-R Interpersonal-Affective factor (rcontrast- CV =.38). These findings suggest a substantial amount of similarity in how the PCL-R and PPI relate to criterion measures despite the differences in their content and method of administration.

11 CLARIFYING PSYCHOPATHY CONCEPTIONS 11 Present Study The primary aim of this study is to clarify which historical conceptualization of psychopathy (Cleckley, 1941; Karpman, 1948a; Lykken, 1957; McCord & McCord, 1964; Mealey, 1995) is most consistent with total and factor scores of the two most widely-researched adult psychopathy measures. Specifically, we used Westen and Rosenthal s (2003) construct validity coefficients to investigate the degree to which the relations between each psychopathy measure and 14 conceptualization -relevant criterion variables. For the PCL-R, we hypothesized that total scores would reflect McCord and McCord s (1964) psychopathy conceptualization, given the PCL-R s focus on impulsivity, aggression, lack of guilt, and callousness and its omission of explicit indicators of anxiety. The McCords underscored not only the importance of lovelessness and guiltlessness in psychopathy, but also impulsivity and aggression. According to them, the psychopath is an asocial, aggressive, highly impulsive person, who feels little or no guilt and is unable to form lasting bonds of affection with other human beings (McCord & McCord, 1964, p. 3), has no stable goals, and is dominated by feeling desires which leave no space for farsighted planning (McCord & McCord, 1964, p. 10). At the scale level, we expected the Interpersonal-Affective factor to reflect primary psychopathy (i.e., Cleckley s 1941, 1976 conceptualization) and the Social Deviance factor to be most consistent with secondary psychopathy (i.e., Karpman s 1941, 1948a conceptualization). For the PPI, we expected total scores to best reflect Lykken s (1957, 1995) primary psychopathy, given the measures substantial emphasis on fearlessness and lack of anxiety (Lilienfeld & Andrews, 1996). At the sub-dimension level, we expected the Fearless Dominance scale to best reflect Lykken s (1957, 1995) primary psychopathy, the Self-Centered Impulsivity

12 CLARIFYING PSYCHOPATHY CONCEPTIONS 12 scale to be most consistent with McCord and McCord s (1964) psychopathy conceptualization given their emphasis on egocentricity and impulsivity, and the Coldheartedness scale to similarly reflect McCord and McCord s conceptualization given that this scale substantially emphasizes lack of guilt and weak emotional attachments to others. Method Participants Drawn from a larger study that used measures of psychopathy to examine heterogeneity in antisocial personality disorder (Poythress, Edens, Skeem, Lilienfeld, Douglas, Frick, Patrick, Epstein, & Wang, 2010), participants comprised a sample of 1,352 male offenders with valid scores on a psychopathy instrument (i.e., PCL-R and/or PPI). Participants were either incarcerated (52.5%) or court-ordered to participate in a community treatment facility (47.5%). Only male offenders were examined for the analyses reported here in view of concerns that psychopathy may manifest differently in females (Cale & Lilienfeld, 2002; Verona & Vitale, 2006). Consistent with previous studies using this dataset (e.g., Poythress, Edens et al., 2010; Poythress, Lilienfeld et al., 2010), data were excluded for 28 participants whose score on the Personality Assessment Inventory s (PAI; Morey, 1991) Inconsistency or Infrequency validity scales suggested invalid responding (i.e., T > 79, Edens & Ruiz, 2005) and for 43 participants who did not complete the PAI or the IQ screen (Quick Test; QT; Ammons & Ammons, 1962). The final sample used in the analyses reported here consisted of 1,281 participants. All were between the ages of 18 and 59 years old (average age = years old; SD = 6.56), competent in English, had an estimated IQ > 70 via the QT, and were not currently prescribed antipsychotic medications. The sample was 64.0% Caucasian, 33.7% African American and 2.3% unidentified.

13 CLARIFYING PSYCHOPATHY CONCEPTIONS 13 Procedure Before participating, all individuals provided informed consent and, in exchange for participation, received $20.00 in compensation (except at one agency that did not allow participant payment). Using IRB-approved procedures, participants were then administered the PCL-R, Structured Clinical Interview for DSM IV Axis II (SCID-II), the Quick Test (QT), and demographic information based on a face-to-face interview and review of file information. Interviewers were graduate-level clinical psychology students with advanced training in the assessment of psychopathy and antisocial personality disorder. Additionally, participants completed the battery of measures described in the following section. Measures Descriptive information for each measure is presented in Table 2. Measures of Psychopathy PCL-R. The PCL-R (Hare, 2003) is a 20-item checklist-based interview completed following a detailed diagnostic interview and review of file information. Each item is rated as 0 (does not apply), 1 (applies somewhat), or 2 (definitely applies) by trained raters. The PCL-R yields a total score and Interpersonal-Affective and Social Deviance scale scores. Before conducting PCL-R interviews, raters completed formal PCL-R training, which included reviewing and scoring several practice videotapes. During the course of the study, the project coordinator made regular site visits to observe the trained raters and complete PCL-R ratings to ensure reliable PCL-R interviewing and scoring. When treating the project coordinator s ratings as the criterion, acceptable levels of inter-rater reliability (ICC =.88) were obtained for 51 cases. PPI. The PPI (Lilienfeld & Andrews, 1996) is a self-report questionnaire that features 187 items scored on a 4-point Likert scale. The PPI contains eight subscales that yield a total

14 CLARIFYING PSYCHOPATHY CONCEPTIONS 14 score, two higher-order, factor scores, Fearless Dominance (FD), and Self-Centered Impulsivity (SCI), along with the standalone dimension of Coldheartedness. FD comprises the Stress Immunity, Social Potency, and Fearlessness scales. SCI comprises the Machiavellian Egocentricity, Impulsive Nonconformity, Blame Externalization, and Carefree Nonplanfulness scales. Measures of Criterion Variables As argued by Campbell and Fiske (1959) the most promising designs for evaluating construct validity feature multi-trait, multi-method designs. Consistent with these recommendations, this study incorporated measures from a variety of methods, in turn designed to detect a variety of individual differences potentially relevant to psychopathy. For ease of presentation and identification, criterion measures are presented in groupings based on their methodology (e.g., self-report, interview). Coding-Based Measures Consistent with Silver (2000), neighborhood disadvantage was coded based on the following 2000 Census tract indices: per capita income (reverse scored), percent of households on public assistance, percent non-white only households, percent of female-headed households, and percent of people unemployed. Laboratory Tasks Affective Deficits. To assess affective deficits, a lexical decision-making task was administered via a laptop computer. Participants were presented a pairing of a word and a nonword on the screen. The actual word was sometimes positive (e.g., sunset, n = 12), negative (e.g., misery, n = 12) and neutral (e.g., bowl, n = 24). In each case, the actual word was paired with a different non-word (n = 48). For each pairing, the participant was asked to identify

15 CLARIFYING PSYCHOPATHY CONCEPTIONS 15 the group of actual words presented on the screen as quickly as possible. Participants completed four blocks of 12 pairings. Research indicates that higher PCL-R scores are related to slower identification of emotional words (Lorenz & Newman, 2002). Passive Avoidance Learning. To detect difficulty with passive-avoidance learning, a deficit long considered central to psychopathy (Lykken, 1995), a GoNoGo Task (GNG; Newman & Kosson, 1986) was administered via a laptop computer. Participants were presented eight 2- digit numbers and asked to learn which four stimuli were associated with punishment (loss of $0.10) and which stimuli were associated with reward (earning $0.10). For each trial, responses were registered during the presentation of a stimulus by pressing a button on a response box. The outcome of interest is the number of times the participant presses the button for punished response. Each participant completed two blocks of 40 trials, with the first block serving as practice and the second as data for analysis. Interview-Based Measures Antisocial History. Based on a diagnostic interview and review of file information, antisocial history was assessed via the antisocial personality disorder (ASPD) module of the Structured Clinical Interview for DSM IV Axis II (SCID-II; First et al., 1997). Total ASPD symptom counts were rendered, which includes both symptoms of Adult ASPD and Conduct Disorder. When treating the project coordinator s ratings as the criterion, acceptable levels of inter-rater reliability (ICC =.86) were obtained for the total symptom counts of 46 cases. Intelligence. To assess intelligence, the Quick Test (QT; Ammons & Ammons, 1962) was administered. In this task, participants are shown a card of four pictures and asked to identify which image best matches the words read aloud by the administrator. Research suggests

16 CLARIFYING PSYCHOPATHY CONCEPTIONS 16 that the QT is strongly correlated with (r =.76) of WAIS-R IQ scores (Craig & Olson, 1988) and intelligence scores in offenders (Doss, Head, Blackburn, & Robertson, 1986; Simon, 1995). Self-Report Measures Abuse. Abuse history was measured with the Child Abuse and Trauma Questionnaire Scales (CATS; Sanders & Giolas, 1991). The questionnaire features 38 items rated on a 5-point scale rating the frequency of events from never to always. Commonly used in studies of childhood abuse (Becker-Lausen, Sanders, & Chinsky, 1995; Ruiz, Pincus, & Ray, 1999), items focus on a negative home environment and verbal, physical, and sexual abuse. Anxiety. The Personality Assessment Inventory s (PAI; Morey, 1991) Anxiety (ANX) scale was developed to index the clinical features of anxiety disorders; the scale focuses on cognitive features such as rumination, affective aspects such as subjective feelings of strain, and physiological signs of tension. Validity of this scale is suggested by large positive correlations with NEO-Personality Inventory (NEO-PI; Costa & McCrae, 1985) Neuroticism (r =.76) and Beck s Anxiety Inventory (Beck & Steer, 1990) total score (r =.62). Behavioral Inhibition System. The Behavioral Inhibition System scale (Carver & White, 1994) is a 7-item scale that gauges how much an individual is prone to inhibiting behaviors that are likely to result in punishment. Validity is indicated by strong associations with measures of anxiety, negative affect and harm avoidance (Carver & White, 1994). BIS was included due to its relevance to Lykken s (1995) primary conceptualization. Egocentricity. The PAI s Antisocial Features Egocentricity scale was developed to assess the egocentric characteristics of antisocial personality disorder and psychopathy. This scale focuses on an individual s self-centeredness, general disregard for others and society, willingness to take advantage of others, and lack of importance placed in relationships with

17 CLARIFYING PSYCHOPATHY CONCEPTIONS 17 spouses and children. Validity is indicated by strong associations with indicators of antisocial personality disorder and psychopathy (Morey, 1991). Fearlessness. The Multidimensional Personality Questionnaire s Harm Avoidance scale (Tellegen, 1982) is a 12-item subscale that gauges how much an individual dislikes danger, disasters and risk, as well as how an individual avoids potential injury. When reverse scored, the measure reflects a propensity toward thrill-seeking and fearlessness. Validity is indicated by moderate negative associations (r = -.36) with Zuckerman s (1979) Sensation Seeking Scale (Patrick, Curtin & Tellegen, 2002). Guiltlessness. The Levenson Self-Report Psychopathy Scale (LSRP) is a 26-item questionnaire that measures psychopathy. This measure includes a 4-item Callousness subscale (Brinkley, Diamond, Magaletta, & Heigel, 2008), whose factor structure has been replicated (Sellbom, 2011) and is used in the analyses reported here. In light of the Callousness subscale s face validity as an indicator of guilt and allied constructs and moderately strong negative association with guilt (r = -.42, p.01; Salekin, Chen, Sellbom, Lester, & MacDougall, 2014), the Callousness subscale is used as a proxy measure for guiltlessness in this study. Impulsivity. The Barratt Impulsivity Scale (Version 11; BIS-11, Barratt, 1994) is a 30- item questionnaire that indexes how an individual displays impulsive tendencies. Validity is supported by significant positive associations with PCL-R Social Deviance factor scores, but not PCL-R Interpersonal-Affective factor scores (Ireland & Archer, 2008). Lovelessness. The PAI Warmth scale assesses the degree to which an individual is empathic, supportive, and warm in interpersonal relationships. Validity is suggested by a moderate association with NEO-PI Extraversion (r =.45) and negative associations with several MMPI personality disorder scales including Avoidant Personality Disorder (r = -.66; Morey,

18 CLARIFYING PSYCHOPATHY CONCEPTIONS ). We reverse-scored the Warmth scale and used it as a proxy measure of lovelessness, regarded by some authors (e.g., McCord & McCord, 1964) as a central feature of psychopathy. Treatment motivation. The PAI Treatment Rejection scale assesses treatment resistance, personal irresponsibility for change, and lack of treatment motivation. Validity is indicated by this scale s predictive utility for non-mutual treatment termination (Hopwood, Ambwani & Morey, 2007) and treatment completion in a forensic sample (Edens & Ruiz, 2005). Analyses Westen and Rosenthal s (2003) construct validity coefficients were developed as an extension of contrast analyses commonly used in analysis of variance comparisons. As noted earlier, these coefficients enable objective comparisons of findings across measures. These metrics (ralerting-cv and rcontrast-cv) may be interpreted like correlation coefficients and summarize the degree to which a measure s observed correlations match its predicted correlations. Consistent with a previous study that used the current sample (Poythress, Lilienfeld et al., 2010), a computational adjustment was needed for rcontrast-cv. Westen and Rosenthal s metric assumes equal n of subjects for all criterion associations. However, different n of subjects were obtained for all criterion variables. To address this issue, the unweighted harmonic mean of n for each conceptualization s set of criterion measures was calculated, as recommended by Robert Rosenthal (see Poythress, Lilienfeld et al., 2010). When accounting for this variation in sample size, rcontrast-cv values decrease along with the unweighted harmonic mean of the sample size, but only slightly. Results Analyses were completed in two stages. First, the predicted associations for each psychopathy conceptualization were generated via a survey of psychopathy researchers. Second,

19 CLARIFYING PSYCHOPATHY CONCEPTIONS 19 Westen and Rosenthal s (2003) ralerting-cv and rcontrast-cv construct validity coefficients were calculated using these survey predictions. Researcher Consensus Predictions Twenty-six researchers with an extensive history of publishing in the field of psychopathy were asked to help develop a survey-informed set of profiles for 7 prominent psychopathy conceptualizations. An effort was made to invite psychopathy researchers with diverse opinions from different research groups to ensure a broad representation of the field. Once selected, researchers were sent an electronic survey that inquired about the importance of 14 constructs to 7 prominent psychopathy conceptualizations. This list of 14 constructs was selected based on their presence in the existing dataset and relevance to different psychopathy conceptualizations. Fourteen 1 of the 26 (54%) researchers completed the survey consistent with responses in previous psychopathy research (n=15; Miller, Lynam, Widiger & Leukefeld, 2001). For each conceptualization and construct, the researcher was asked two questions. First, the researcher was asked to indicate (yes/no) whether the construct is an important characteristic of a specific conceptualization of psychopathy. For example, if the researcher believed that fearlessness is an important characteristic of Lykken s (1995) conception, he or she would be asked to enter yes. Second, for constructs that the researcher identified as important to a conceptualization, the researcher was asked how this characteristic should correlate with a conceptualization. The researcher was given seven options for rating the correlation: strong negative association, moderate negative association, weak negative association, negligible association, weak positive association, moderate positive association, and strong positive association. For example, if the researcher believed that Lykken considered a lack of fear to be

20 CLARIFYING PSYCHOPATHY CONCEPTIONS 20 the trait that explained most or all other psychopathy-related features, he or would be asked to enter strong positive association for fearlessness. These survey ratings were converted into psychopathy conceptualization prototypes for use in construct validity coefficient analyses in two steps. First, a construct was included in the prototype for a given conceptualization if at least two-thirds of respondents identified that construct as important to a conceptualization. This cut-off was selected to balance the need for respondent agreement and distinction among psychopathy conceptualization prototypes. Second, consistent with other research using construct validity coefficients (Dibble, Hartmann, & Rosaen, 2016), we assigned negligible, weak, moderate, and strong associations values of.00,.10,.30, and.50, respectively (Cohen, 1988). In turn, we averaged these values to obtain the hypothesized effect size for a construct (see Table 3). There are several important divergences between survey responses and the writings of theorists. These discrepancies might have arisen because of some researchers relative lack of familiarity with certain psychopathy conceptualizations or because of the survey s wording. For example, in aggregate, the researchers surprisingly did not consider anxiety an important part of Karpman s conceptualization of secondary psychopathy. This could be because the survey listed the construct as Low Anxiety. It may be that researchers would have considered High Anxiety as relevant to Karpman s secondary psychopathy, but did not realize that they could predict a strongly negative association for Low Anxiety. To address this and similar issues, the three co-authors of this study and a collaborator (Christopher Patrick) also completed the survey (Table 4). The agreement between the researchers surveyed and the co-authors was calculated with Kappa. Using the 2/3rds cut-off approach, constructs not viewed as important for a conceptualization were coded as 0 and

21 CLARIFYING PSYCHOPATHY CONCEPTIONS 21 constructs viewed as important for a conceptualization were coded as a 1. The Kappa for all predictions was.51. Generally, Kappa values of less than.40 are considered poor, are fair, are good, and.75 and above are excellent (Cichetti & Sparrow, 1981). Agreement between researchers and co-authors fell in the "fair" range, indicating that these two groups generally agreed but there are some divergences. To resolve inconsistencies between the researcher responses and the source material of psychopathy theorists, we supplemented the researcher responses when at least two-thirds of the co-authors identified a construct as important to a conceptualization. Using the example earlier, less than two-thirds of the researcher survey respondents made a prediction regarding anxiety for Karpman's secondary conceptualization whereas all four of the co-authors did. In this instance, we added the co-author s prediction to the researcher psychopathy conceptualization prototype for Karpman s secondary conceptualization. There were a total of 16 changes (16.3%) made out of 98 opportunities (i.e., 14 constructs for each of 7 conceptualizations) when requiring the support of at least two-thirds of the co-authors to supplement the researchert responses. Thus, responses are heavily skewed toward researcher ratings. The final psychopathy conceptualization prototypes are presented in Table 5. Construct Validity Coefficients After the researcher consensus predictions were finalized, Westen and Rosenthal s (2003) construct validity coefficients were calculated to identify what conceptualizations are consistent with PCL-R and PPI conceptualizations of psychopathy. Because rcontrast-cv indexes not only how well the researcher predicts the observed set of associations (see Table 6), but also how well the correlations explain the variance of a given measure, we focused our primary analyses on this index. For each measure, these rcontrast-cv findings were examined in two ways. First, the relative

22 CLARIFYING PSYCHOPATHY CONCEPTIONS 22 fit was identified based on the strongest rcontrast-cv value (See Table 7). Second, the absolute fit of each measure (see Table 8) was identified by computing the average rcontrast-cv value across primary conceptualizations (i.e., Cleckley, Karpman, Lykken and Mealey), and secondary conceptualizations (i.e., Karpman and Mealey). This provides a broad index of the consistency of a given measure with the tested conceptualizations. Although rcontrast-cv findings are emphasized, ralerting-cv was computed as a supplementary analysis and can be viewed in Table 7. PCL-R. The PCL-R s total score was moderately consistent with the McCords conception (rcontrast-cv =.36), which emphasizes lovelessness and guiltlessness, and Karpman s affective dysfunction-focused secondary conception (rcontrast-cv =.36). At the subscale level, the PCL-R Social Deviance factor evinced a pattern of results similar to the PCL-R total score. Specifically, the PCL-R Social Deviance factor demonstrated strong consistency with the McCords (rcontrast-cv =.50) and Karpman s secondary (rcontrast-cv =.51) conceptualizations of psychopathy. In contrast, the PCL-R Interpersonal-Affective factor was weakly consistent with Lykken s fearlessness-based (rcontrast-cv =.22) and Cleckley s affective deficit-focused (rcontrast-cv =.21) conceptualizations of primary psychopathy. Findings at the absolute fit level were largely consistent with these findings. Here, the PCL-R total score and Social Deviance factor were moderately to strongly consistent with conceptualizations of secondary psychopathy, whereas the PCL-R Interpersonal-Affective scale only demonstrated weak consistency with primary psychopathy conceptualizations. PPI. The PPI s total score was strongly consistent with the McCords (rcontrast-cv =.52), Karpman s secondary (rcontrast-cv =.48), and Mealey s secondary (rcontrast-cv =.46) conceptualizations of psychopathy. These findings generalized from the PPI total score to the PPI Self-Centered Impulsivity subscale, which demonstrated even stronger coefficients for the

23 CLARIFYING PSYCHOPATHY CONCEPTIONS 23 McCords (rcontrast-cv =.77), Karpman s secondary (rcontrast-cv =.77), and Mealey s secondary (rcontrast-cv =.74) conceptualizations of psychopathy. In contrast, the PPI Fearless Dominance and Coldheartedness subscales exhibited moderate to strong coefficients for Lykken s fearlessnessbased (Fearless Dominance rcontrast-cv =.54; Coldheartedness rcontrast-cv =.42) and Cleckley s affective deficits-focused (Fearless Dominance rcontrast-cv =.48; Coldheartedness rcontrast-cv =.49) conceptualizations of primary psychopathy. Consistent with these findings at the relative fit level, absolute fit estimates reveal that both the PPI total score and Self-Centered Impulsivity subscale were more consistent with secondary psychopathy theories, whereas the Fearless Dominance and Coldheartedness scales were better characterized by primary psychopathy theories. Discussion The overarching goal of this study was to identify the conceptualizations of psychopathy that are best reflected by the two most widely researched adult psychopathy measures. Rather than indicating that the PCL-R and PPI measure markedly different conceptualizations as we hypothesized, our findings suggest considerable convergence between these measures. First, at the total score level, both the PCL-R and PPI were most consistent with conceptualizations of secondary psychopathy, which emphasize affective disturbance and dysfunction. Second, within measures, scales of interpersonal traits and affective deficits (i.e., PCL-R Interpersonal Affective and PPI Fearless Dominance scales) were most consistent with primary psychopathy conceptualizations (i.e., Lykken) whereas measures of social deviance (i.e., PCL-R Social Deviance and PPI Self-Centered Impulsivity) better indexed secondary psychopathy conceptualizations. Nonetheless, perhaps surprisingly, the PCL-R generally demonstrated construct validity coefficients that were weaker in magnitude relative to those of the PPI.

24 CLARIFYING PSYCHOPATHY CONCEPTIONS 24 Limitations Although this study was marked by several strengths, including a large sample size, multi-method measures, and a longitudinal design, its limitations warrant careful consideration. First, although the 14 criterion measures included in this study are important, they are not exhaustive. Adding other criterion measures (e.g., behavioral activation system strength; a more direct measure of lovelessness) could have altered the findings. Second, some criterion measures (i.e., abuse, affective deficits, passive avoidance learning, and poverty) did not manifest predicted associations with the PCL-R or PPI scores, which may have limited the magnitude of rcontrast-cv coefficients. This concern is partially alleviated by the fact that the magnitudes of associations observed here are broadly within the range reported in previous research. Third, despite efforts to limit mono-operation bias in criterion measure selection, many measures were still self-report (64%). This high proportion of self-report measures may have contributed to higher construct validity coefficients, particularly rcontrast-cv, for the PPI, which is self-report, in comparison with the PCL-R, which is interview-and file-based. To address this study s limitations, the present findings should be replicated in research with rich datasets that include a more balanced set of measures. Key Findings: Overall There was a great deal of consistency in which theories best matched the PCL-R and PPI. We found that the PCL-R and PPI total scores were most consistent with the McCords conception of psychopathy and Karpman s secondary psychopathy. Similar to total scores, the PCL-R Social Deviance and PPI Self-Centered Impulsivity were most consistent with the McCords conception of psychopathy and Karpman s secondary psychopathy. In contrast, the PCL-R Interpersonal Affective, PPI Fearless Dominance, and PPI Coldheartedness scales were

25 CLARIFYING PSYCHOPATHY CONCEPTIONS 25 most consistent with Lykken s and Cleckley s primary psychopathy. The lack of convergence of the PPI Coldheartedness scale with the McCords conceptualization of psychopathy was unexpected and warrants further research, although it may in part reflect the absence of explicit indices of guiltlessness and lovelessness the two features most emphasized by the McCords - in our criterion measures. This pattern of findings is also surprising in light of the key divergences between these two major psychopathy measures. The PCL-R excludes explicit indicators of positive adjustment (e.g., good intelligence, absence of nervousness, and suicide rarely carried out), whereas the PPI explicitly assesses a lack of trait anxiety and interpersonal dominance/self-confidence. The PCL- R includes both items that explicitly gauge antisocial or even criminal behavior and items that can be scored on the basis of criminal acts, whereas the PPI explicitly omits such content. These differences have led some to wonder if the PCL-R and PPI identify overlapping but separable constructs. In hindsight, these findings were loosely foreshadowed by those of Poythress, Lilienfeld et al. (2010), who found a great deal of similarity between how the PCL-R and PPI total scores correlated with key criterion variables. At the subscale level, the PPI Fearless Dominance scale was most consistent with the PCL-R Interpersonal-Affective scale, whereas the PPI Self- Centered Impulsivity scale was most consistent with the PCL-R Social Deviance scale (Poythress, Lilienfeld et al., 2010). This was despite the fact that the scales themselves were correlated at weaker levels than one would expect for measures of the same construct (Poythress, Lilienfeld et al., 2010). Findings from these studies suggest that the PCL-R and PPI relate to most criterion variables (Poythress, Lilienfeld et al., 2010) and predicted patterns of associations

26 CLARIFYING PSYCHOPATHY CONCEPTIONS 26 with criterion measures (the present study) similarly, despite differences in item content and minimal convergence in inter-correlations among these psychopathy measures. Key Findings: PCL-R Despite similarities in which models are most consistent with the PCL-R and PPI, one key distinction is that the PCL-R demonstrated lower construct validity coefficients with these theories than the PPI. This was particularly true of primary psychopathy theories. For example, the PCL-R Interpersonal-Affective scale (rcontrast-cv =.21, CIs =.16,.26) demonstrated a weaker rcontrast-cv value than the PPI Fearless Dominance scale (rcontrast-cv =.42, CIs =.37,.47). The PCL-R s limited consistency with primary theories of psychopathy may be unexpected given the degree of overlap between Cleckley s primary psychopathy criteria and the PCL-R, particularly in terms of affective deficits (Patrick, 2006). Nonetheless, Hare significantly diverged from Cleckley s criteria during the development of the PCL and later, the PCL-R, in part by omitting indicators of positive adjustment and including items focusing on antisocial behavior. This decision may have in part stemmed from the fact that PCL was also informed by the writings of Karpman, the McCords, and other scholars (Hare & Neumann, 2010). As a result, despite the PCL-R s distinctively Cleckleyan roots, it may not map directly onto only one historical conceptualization of psychopathy. Key Findings: PPI Each PPI scale and total score was strongly consistent with historical conceptualizations of psychopathy. In fact, the PPI total score was moderately to strongly consistent with four different conceptualizations, including the McCords, Lykken s primary, and both Karpman s and Mealey s secondary. At the subscale level, strength of consistency was more apparent for the PPI Self-Centered Impulsivity scale than Fearless Dominance and Coldheartedness scales.